VSE knjižnice (vzajemna bibliografsko-kataložna baza podatkov COBIB.SI)
  • Feature extraction methods efficiency
    Dreo, Dejan
    In this session the influence of feature extraction methods on speech recognition rate will be presented. The features in focus were linear prediction features, cepstral features, mel cepstral ... features and compound features combined with linear prediction features, delta linear prediction features, mel cepstral features, delta mel cepstral features and energy. In case of linear prediction, cepstral and mel cepstral features the size 12, 24 and 36 elements per feature vector were used. In case of compound feature vector the 12 elements linear prediction, 12 elements delta linear prediction, 12 elements mel cepstral, 12 elements delta cepstral feature vector were used and energy of the frame added.
    Vir: Advances in speech technology : proceedings (Str. [39-49])
    Vrsta gradiva - prispevek na konferenci
    Leto - 1998
    Jezik - angleški
    COBISS.SI-ID - 3939350

Vnos na polico